KR0146995B1 - Method for amendment and restroration of replicated data of client server structure in atm exchanger - Google Patents

Abstract

The present invention relates to a method of changing and recovering redundant data having a client-server structure in an asynchronous delivery mode switch, wherein some processors having redundant data allow real-time change of data even in an abnormal or communication failure. In order to provide a redundant data change and recovery method having a fault-tolerant client-server structure that allows recovery to the same data as the processor, when duplicate data change is requested in the standby state, the redundant data of the OMP 24 is changed and the disk 25 A first step (31 to 35) broadcasting the change information after changing the redundant data of the " A second step (36, 37) of receiving a change complete signal from each processor, classifying and processing an abnormal processor, and then transitioning to the standby state; A third step (31) of retrieving the disk 25, extracting duplicate data information, restoring a directory and dictionary, restoring the duplicate data, and then transitioning to the standby state when the duplicate data recovery is requested in the standby state. 38 to 42), real-time response is possible, the availability of data is high, and there is an effect that can solve the trade-off relationship between the guarantee of data consistency and performance to some extent.

Description

Redundant Data Change and Recovery Method with Client-Server Structure in Asynchronous Delivery Mode

1 is a structure of a distributed real-time DBMS in ATM interleaver according to the present invention,

2 is a block diagram showing the configuration and signal flow between blocks for a fault-tolerant client-server structure for changing and recovering redundant data;

3 is a general flow diagram of redundant data change and recovery in accordance with the present invention.

* Explanation of symbols for main parts of the drawings

21,22,23: SCP 24: OMP

25: disk

The present invention relates to a redundant data change and recovery method having a client-server structure in an asynchronous delivery mode switch.

In most of the conventional distributed data base management system (DBMS), the change of the duplicate data is performed in a batch by a transaction or in a store-and-forward manner.

Therefore, in the former case, some processors cannot be changed during a failure or communication failure, and in the latter case, a time delay occurs in matching data. In addition, since all processors need to access the disk to change the redundant data, the time-consuming and time-consuming real-time response is excessive. In addition, the distributed real-time DBMS used in the traditional interlining machine also has a problem that the change request itself is rejected and the application must be performed again when some processors cannot be changed due to an inter-processor communication failure or an abnormality of a transmission signal.

In order to solve the above problems, the present invention allows some processors with redundant data to be changed in real time even when abnormal or communication failure occurs, and a fault tolerance client which recovers to the same data as other processors when the processor is restarted. Its purpose is to provide a redundant data change and recovery method that has a client-server structure.

In general, data consistency and real-time response are known to be trade-off. However, in the present invention, data changes overlapping two or more processors are made in real time even in an environment of a failure or communication failure of some processors, and guarantees consistency of duplicate data when restarting a specific processor. In other words, the server for Replicated Date (SRD) has the characteristics of fault tolerance and real-time response.

In the present invention for achieving the above object, in the redundant data change and recovery method applied to the asynchronous transfer mode (ATM) switch, if the duplicate data change is requested in the standby state, the redundant data of the OMP is changed and the OMP duplicate data of the disk A first step of broadcasting the change information after the change; A second step of receiving a change complete signal from each processor after performing the first step, classifying and processing an abnormal processor, and then transitioning to the standby state; And a third step of recovering a directory and a dictionary, restoring a directory and a dictionary, restoring the redundant data, and then transitioning to the standby state when the redundant data recovery is requested in the standby state. do.

Hereinafter, with reference to the accompanying drawings will be described an embodiment according to the present invention;

1 is a distributed real-time DBMS structure diagram of an ATM switch system to which the present invention is applied.

The Asynchronous Transfer Mode (ATM) exchange consists of a subsystem of the Access Central Subsystem (ACS) and the Acts Local Subsystem (ALS). ACS manages the entire system and provides operator matching.

The ACS consists of an Operation and Maintenance Processor (OMP), an operator matching device, a data storage device and a Central SWitch (CSW).

The ALS is composed of a subscriber call processor (SCP) that manages various connection states and a state of a hardware device, a subscriber matching device, and an access switch (ASW).

OMP connects to one port of the CSW and SCP connects to one port of the ASW via an Inter-Module Interface (IMI) link. The IPC path between SCP and SCP is a three-stage structure of ASW-CSW-ASW, and the communication between OMP and SCP is a two-stage structure of CSW-ASW.

There is a DBMS that includes DBBG (Data Base Backup Group) in OMP of ACS for fault-tolerant change and recovery of redundant data. OMP's DBMS consists of DBBG, Data Base Supporting Group (DBSG), Data Base Query Group (DBQG), Data Base Transcation Group (DBTG), and Data Base Kernel Group (DBKG) blocks. SCP's DBMS consists of DBKG, DBSG, and DBTG.

The DBSG performs the communication between the user registration function and the remote processor DBMS. DBBG performs the change control of duplicate data, and DBKG performs the function of searching, changing, inserting and deleting the actual database. DBQG handles query terms, and DBTG handles transaction processing.

FIG. 2 is a diagram illustrating the configuration and signal flow between related blocks for implementing a fault-tolerant client-server structure for changing and recovering redundant data in an ATM switch.

Each processor has its own local DB (LDB) in main memory, and the same DB is copied to OMP disk. Client for Replicated Data (CRD), a client of duplicate data existing in DBSG of SCP1 (22) where duplicate data has changed, finds the relative distance from the relation to be changed in the relation to be changed through DBKG, and DBBG of OMP 24 Control is requested from SRD, a duplicate data server existing in a block.

The SRD modifies the duplicate data in the OMP memory LDB and the disk's OMP duplicate data and, if successful, allows SCP1's CRD to duplicate the duplicate data. At the same time, the SRD searches for processors with redundant data in the LDB and broadcasts a change request of the redundant data to the CRDs of the related processors.

The CRD of each processor receives a duplicate data change request from the SRD, changes the LDB of the main memory, and responds to the SRD. The SRD waits for some time and receives a change completion signal from the DBSG of each processor. In this process, SRD analyzes each processor's duplicate data change status to find out which processor failed or did not respond, and completes the duplicate data change. Abnormal processors are stored in the disk 25 if the processor is not being recovered and information about the change of the redundant data. If a bad processor is recovering for restart, it will be temporarily stored in the local buffer of the SRD. At this point, the redundant data on the disk and the main memory of the OMP are consistent, but the remaining processors are not consistent because the disk data has old values. This mismatch is not a problem as long as the processors are in normal operation. However, when restarting, incorrect data is loaded and such inconsistency is resolved in the process of recovering duplicate data. Abnormal processors are monitored periodically, and when it recovers to normal, the change information is transmitted and changed to maintain the consistency of duplicate data.

Therefore, it is possible to prevent the change of the duplicate data to be rejected or delayed due to an abnormal state of a specific processor, thereby greatly increasing the flexibility of real-time processing and data change compared to other conventional exchanges.

The recovery of duplicate data is divided into OMP and non OMP. First, when the OMP 24 is restarted, since the main memory and the disk 25 are changed at the same time in the process of changing redundant data, the disk DB may be loaded. In the case of processors other than OMP, only the main memory is changed when the redundant data changes, so the disk DB must be restored before loading. P_REP in OMP disk is a file containing information about duplicate data. The number and name of duplicate relations between OMP, the master processor of duplicate data, and the rest of the processors containing duplicate data. Dictionary addresses, starting addresses of duplicate data, and so on, define the association.

The SRD copies redundant data of the restarted processor by referring to the redundant data information from the disk DB of the OMP. In addition, the recovery of the duplicated data is completed by applying the logs of the local buffer of the SRD in which the duplicated data generated during the disk copy and not applied to the OMP disk DB are stored.

3 is a general flow diagram of SRD, which is a redundant data server of an ATM switch distributed real-time DBMS.

The redundant data server is in a standby state (31) and receives a request for changing redundant data (32) from the CRD or a request for recovering redundant data (38) from the system loader. Upon receiving a request for changing the redundant data from the DBSG, the redundant data of the LDB in the main memory of the OMP is changed (33) and the OMP copy DB on the disk is changed (34). When the change of the disk data is completed, all processors including the duplicate data are broadcasted to change the duplicate data (35). After a certain time, a change completion signal is received from each processor (36), the abnormal processor is classified and processed (37), and waits to process the next request.

When the redundant data server receives a request for recovery of the duplicate data from the system loader (38), the duplicate data existing in the OMP disk is searched (39) to extract all duplicate data information of the restarted processor. Copy and recover the directory and dictionary for duplicate data from the directory and dictionary of OMP DB (40) and complete the disk copy for duplicate data by copying the actual data from OMP DB. In order to reflect the changes of the duplicate data that occurred during the disk copying of the North and North Korea data, the log data in the local buffer of the SRD is applied and the recovery completion is sent to the system loader to complete the recovery of the duplicate data (41, 42). . It waits to process the next request.

The present invention as described above is effective as follows.

First, real-time response is possible by minimizing disk access in the process of changing redundant data.

Second, the availability of data is high because it allows communication failures or failure of some processors.

Third, it is possible to solve some of the teide-off relationship between guaranteeing the consistency of data and performance.

Claims (1)

In the redundant data change and recovery method applied to an asynchronous transfer mode (ATM) exchange, if a redundant data change is requested in the standby state, the redundant data of the OMP 24 is changed and the OMP redundant data of the disk 25 is changed. A first step (31 to 35) of broadcasting change information; A second step (37, 37) of receiving the change complete signal from each processor after performing the first step (31 to 35), classifying the abnormal processor, and processing the transition to the standby state; A third step (31) of retrieving the disk 25, extracting duplicate data information, restoring a directory and dictionary, restoring the duplicate data, and then transitioning to the standby state when the duplicate data recovery is requested in the standby state. 38 to 42), characterized in that the redundant data change and recovery method.